You are on page 1of 8

Food Quality and Preference 30 (2013) 320–327

Contents lists available at SciVerse ScienceDirect

Food Quality and Preference
journal homepage: www.elsevier.com/locate/foodqual

Colour training and colour differences thresholds in orange juice
Rocío Fernández-Vázquez a, Carla M. Stinco a, Dolores Hernanz b, Francisco J. Heredia a, Isabel M. Vicario a,⇑
a b

Food Colour & Quality Laboratory, Department of Nutrition & Food Science, Universidad de Sevilla, Facultad de Farmacia, 41012 Sevilla, Spain1 Department Analytical Chemistry, Universidad de Sevilla, Facultad de Farmacia, 41012 Sevilla, Spain

a r t i c l e

i n f o

a b s t r a c t
This study was aimed at training a panel of assessors to evaluate specifically orange juice colour, and to establish the colour difference threshold in orange juice for a trained and untrained panel. Panellists were first preselected using Farnsworth–Munsell 100-Hue Test and then trained with a specific method for orange juice colour. This training allowed assessors to evaluate visually orange juice samples in hue and intensity. The final selection of assessors was a panel of 8 trained observers with reproducibility and repeatability, and a significant discrimination among the samples (p < 0.05). On the other hand, commercial orange juices were evaluated both instrumentally by image analysis and visually by the trained panel, and the untrained panel. Instrumental colour measurements and visual evaluation were correlated. Values around 1.5 and 2.8 CIELAB units could be consider the threshold for colour differences between two orange juices for the trained and untrained panel, respectively. Ó 2013 Elsevier Ltd. All rights reserved.

Article history: Received 1 October 2012 Received in revised form 3 April 2013 Accepted 31 May 2013 Available online 12 July 2013 Keywords: Colour Orange juice Sensory training Colour differences

1. Introduction Colour is one of the most important visual attributes in food and usually is the first one evaluated by consumers and is associated to the concept of quality (Huggart, Petrus, & Buzz Lig, 1977; Pangborn, 1960; Tepper, 1993). In orange juices, the natural bright colour is considered one of their main advantages over other juices (Barron, Maraulja, & Huggart, 1967) and has attached great importance since some studies have probed that it may influence flavour perception and other quality attributes (FernándezVázquez et al., 2012; Tepper, 1993). Colour can be evaluated by instrumental or visual analysis. Humans and instruments measure colour in different ways. Human perception of colour is based on responses of photoreceptors in the retina of the eye and the way they are interpreted within the brain. These perceived colours are often characterised by physical scientists using three dimensions: lightness, hue and chroma. Instruments, on the other hand, are capable of seeing pure values of the colorimetric coordinates CIELAB L⁄, a⁄, and b⁄. Nowadays, there are new advances in image acquisition technology that offer the possibility of using technically sophisticated apparatus available at relatively low cost to evaluate colour in terms of millions of pixels. In comparison with the traditional light sensors, the main advantage is that they allow making a detailed evaluation of a wider area of any food product, with inhomogeneous colour possible. Every different colour in the image of the analysed food
⇑ Corresponding author. Tel.: +34 954556339; fax: +34 954557017.
1

E-mail address: vicario@us.es (I.M. Vicario). http://www.color.us.es.

matrix can be accounted for by one or more pixels (Antonelli et al., 2004). Furthermore, it is based upon digital cameras, which can quickly capture images in digital format (DigiEyeÒ) (Luo, Cui, & Li, 2001) and offers a more reliable measurement of the food colour, which can be correlated with sensory analysis and other colour measurements (Fernández-Vázquez, Stinco, MelendezMartínez, Heredia, & Vicario, 2011). Anyway, colour measurement usually requires instruments that are not always available in small and medium size companies and visual evaluation could be an alternative. Human colour vision is a quite complex process and colour is undoubtedly a perception, a virtual property of the material. In order to use the visual analysis as an objective quality control, it is necessary to standardize the measurement conditions to be able to compare with the instrumental measurement. Previous studies have shown that a good correlation can be achieved when the instrumental and sensory measurements are done considering different aspect such as background, surround or illumination (Fernández-Vázquez et al., 2011; Meléndez-Martínez, Vicario, & Heredia, 2005). Although colour evaluation is included in many sensory studies (Calvo, Salvador, & Fiszman, 2001; Frata, Valim, & Monteiro, 2006; Poelman & Delahunty, 2011), there are very few studies which specially train the panellists to do the visual evaluation of food with more details. An example of an specific training in visual evaluation was done by Gambaro, Giménez, and Burgueño (2001) for strawberry yoghourt. Based on this experience, we have particularly trained a panel to evaluate orange juice colour in a reproducible and repeatable way. On the other hand, the evaluation of colour differences has had a high interest for long time. Specifically, ‘just noticeable

0950-3293/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.foodqual.2013.05.018

For objective colour specifications. (2001). Instrumental colour measurement The images digital Tokyo. & Hita. and is used to define the difference of a colour with reference to a grey colour with the same lightness. At the present time. or in the study of the colour fading in food. etc. and Negueruela (2001).2. based on computational solutions (León. Then. 1. However. Domingo. the samples were placed in 75 mL capacity transparent plastic bottles (Fig. Melgosa. & Balonon-Rosen. Pérez. 2006). works of art. or more recently to determine colour tolerance in orange juice (Wei.R. Pérez. This attribute is related to the differences in reflectance at different wavelengths and is considered the qualitative attribute of colour. a⁄ and b⁄: DEà ab ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi à 2 2 ðDLÃ Þ þ ðDaà Þ2 þ ðDb Þ 2. a protocol was designed for the selection and training of assessors based on the methodology proposed by Gambaro et al.. Ou. such as the reproducibility of colour in manufactured products and communication systems. the samples were illuminated by a diffuse D65 simulator. Colour training Briefly. literature on orange juice colour does not provide data on the colour differences that can be visually detected between two orange juices (based on real samples) by consumers. Scheme of the DigiEye System. etc. The objectives of this study were: (1) to train a panel of assessors to evaluate specifically the orange juice colour. In these measurements. those panellists who did not present a good skill to discriminate light differences in tone and intensity using blended of colouring dilutions were 4 1 5 3 2 2 1: Domed cabinet 2: Fluorescent tubes. (2012) established the colour of an ideal orange juice and the colour tolerance. Fernández-Vázquez et al. we used the DigiFood software (Heredia.2-megapixel Nikon D80 (Nikon Corporation. which allows the transformation of RGB values into the CIELAB colour parameters. using a digital display. Previous studies have explored colour threshold using colour standards (Berns. An attempt to stablish the thresholds for visual discrimation between wines was also published by Martínez. González-Miret. Huertas. 2) and measured against a grey surround (L⁄ = 50) and white background. a preliminary panellist selection was made using the Farnsworth-Munsell 100-Hue Test. and an illumination box designed by VeriVide Ltd. 2001). The latter system includes a calibrated camera 10. Hita. Yebra. and (2) to study the visually perceived colour difference by the observers’ panel (trained and untrained) in a complete range of orange juices of different colours to establish the colour difference threshold in this popular beverage. Material and methods 2. the psychophysical parameters chroma (C à ab ) and hue (hab) are defined as: Cà ab ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi à 2 à ðaà Þ2 þ ðb Þ . First. Japan) and an objective Nikkor 35-mm f/2D (Nikon Corporation). For obtaining CIELAB coordinates. and is considered the quantitative attribute of colourfulness. & Ramírez. 2006). 1).Colour differences (DEà ab ) were calculated as the Euclidean distance between two points in the 3-D space defined by L⁄. / Food Quality and Preference 30 (2013) 320–327 321 differences’ have been very important in the development of the colorimetry.1. & León. Álvarez. D65 simulator 3: Sample 4: Digital camera 5: PC Digifood® Software Fig. greenish. 2012). & Hutchings. DigiEye imaging system was used to capture the digital (Luo et al. Franco. 2001). Reniff. Hue (hab) is the attribute according to which colours are usually defined as reddish. 2. a colour sensor for display calibration. UK) (Fig. Alman. so far. One of the key problems in the visual evaluation is establishing the threshold for colour differences. . 1991). Snyder. Luo. although this type of information could be very useful for the orange juice industry. hab ¼ arctan ðb =aÃ Þ Chroma (C à ab ) is used to determine the degree of difference of a hue in comparison with a grey colour with the same lightness. Digital images were made in order to obtain the total appearance of juice at depths observed by consumers. This is the positive number which stays invariable when the products are exchanged (Melgosa. From the CIELAB uniform colour space. (Leicester. Recently Wei et al. which are mathematical expressions which allows us to obtain the number DEà ab . calculation of colour differences has many applications in colorimetry. There are equations to find out the colour difference between two stimuli in the CIELAB space.

chroma and lightness were not considered separately as individual attributes because in previous studies it was observed that panellists had difficulties to understand and evaluate chroma (FernándezVázquez et al. Hue was evaluated from yellowish to reddish and intensity was evaluated from low to high.3. were used to train the assessors in colour evaluation. 2012). sample. being the order randomised across assessors. This design allowed training of the assessors in evaluating orange colour intensity and hue as well as determining their reproducibility and performance consistency. Two commercials orange juice samples (COJ I and COJ II). and they were asked to rank them according to the increasing intensity. According to the results obtained. 2. The colour attributes trained were hue and intensity. 2. the assessors were given the whole dilution series (yellowish. and 3. Sensory evaluation The samples were compared by pairs (120 comparisons) by the trained panel of 8 observers with normal colour vision and previously trained in colour discrimination experiments.322 R. Ten aqueous solutions of each blend of colourings were prepared (100. To select the assessors. and 280 mm depth) to control illumination and observation conditions. the reader is referred to the web version of this article. aqueous orange-coloured solutions were prepared using two food dyes (red and yellow food dyes from McCormick. A collection of samples were selected to encompass the full range of colour intensity and hue in commercial orange juice. (For interpretation of the references to colour in this figure legend. and six samples prepared from dilutions of COJ II (6%. 30%. 22. Observers were situated 50 cm in front of the samples.1. 36. In this study. 29. assessors evaluated duplicates of the commercial samples and a couple of diluted samples. orangish. repeated the experiment with the aim of establishing the colour difference threshold for untrained observers. 50%. Spain S. Colour differences thresholds 2. a white background and a grey surround were selected to simulate the objective measurements made by image analyses (Stinco et al.2. 2011). The samples evaluated in each session were the same for all assessors. intensity was assayed as the best way to evaluate both parameters visually. 17. 2007). They did the test twice per pair of samples in two different sessions: one with the couple of samples 74 mm 29 mm 48 mm Fig. The test was carried out using a VeriVide CAC Portable cabinet (dimension of viewing area: 635 mm width. eight panellists recruited also from students and staff at the University of Seville. In a second stage (intensity ranking). Afterwards. 6 from squeezed oranges and stored at 4 °C.2. 10%.. 1986). Orange juices samples 16 commercial orange juices (5 from concentrated. with white background and grey surround. those assessors with the highest variance and greater judgement dispersion were withdrawn from the panel. The task of each observer was to judge whether they could notice the colour difference between the two orange juice samples (1) or if they could not (0).000 lL/L of red and yellow food dye for orangish dilutions. and to measure the zones of colour confusion in colour defective persons. Finally. 2.1. 60%.000 lL/L of red and yellow food dye for reddish dilutions.2. A three factor ANOVA (assessor. anchored at the end. 1. D65 was used as source of illumination (the same used in the instrumental measurements) (CIE. To determine the ability to discriminate among slight tone and intensity differences in orange colour.250 and 12. In each session. Panel training Two nonstructured 10 cm long scales. and at the end of each session a meeting of 30 min was done by the leader of the panel and all the panellists to unify the criteria of evaluation. 78. orangish and reddish) separately.000 and 12. the selected panel was training using two commercial samples and a serie of dilutions of one of them. Characteristics of the bottles used for containing the samples in the instrumental colour measurements and the visual evaluation.A) blended in different proportions: 500 and 12. and reddish hues (10 evaluations). The FarnsworthMunsell 100-Hue Test for the examination of Colour Discrimination (Farnsworth 1957) was used to verify the normal vision.3. These solutions (75 mL) were placed in bottles of transparent plastic and coded with 2 digit random numbers.000 lL/L of red and yellow food dye respectively. Fernández-Vázquez et al.. For this reason. and 80%) were used. 10. 2. for yellowish dilutions.05) (O’Mahony. 280 mm height. with normal vision (according to the Farnsworth-Munsell 100-Hue Test) but no previous knowledge in colour science. 2. These samples were chosen in order to collect the variety of the orange juices colour available in the supermarket. These samples were evaluated in 15 min sessions. Panel selection A panel of 12 assessors were recruited from students and staff at the University of Seville and then preselected according to their normal colour vision following ISO 11037 (1999). and 5 from squeezed oranges and stored at room temperature) were purchased from different supermarkets in Spain.2. It allows to separate persons with normal colour vision into classes of superior.) rejected. 120 pairs of samples were displayed on the centre of the VeriVide CAC 120 cabinet. and 5%).3. and repetition) for samples COJ I and COJ II. 47. and a two factor ANOVA (assessor and sample) for these two samples and all the dilutions were performed on the data obtained (O’Mahony 1986). the criteria used were: (1) reject those who were unable to accomplish sorting the tubes into the three . 14. average and low colour discrimination. Each sample was placed in 75 mL capacity transparent plastic bottles to measure its colour by image analyses and then to evaluate the colour differences. / Food Quality and Preference 30 (2013) 320–327 tone groups and (2) reject those whose Spearman’s ranked correlation coefficients (p) of sensory ranking versus colour concentrations were not significant (p > 0. The evaluation sessions were organised as follows: in a first stage (hue classification) assessors dealt with the samples (n = 3) corresponding to the each dilution level separately and were asked to sort them in yellowish.

were analysed by image analysis.48 for the yellowish.12–60. 46. and another one with the samples separated by 15 cm (experiment b). hue ranged from 78. 2007).5 (the most reddish OJ) to 92. & Montag.58 in hue and intensity. 1953. Massachusetts) was used for this purpose.45–65.4. Natik. an S-shaped curve (y = A/[1 + exp(B + Cx)]) was fitted using an iterative algorithm of successive approximations to the function and its derivatives.23. 3.5% in intensity (both cases less than 10%) in agreement with an increasing uniformity of the panel.51 to 70.3° in the orangish serie and from 30. Panel selection The Farnsworth-Munsell 100-Hue Test was applied to the assessors and results showed that all passed the test with punctuation lower than 48. moderate cases would show small ‘bulges’ and lower total error scores. / Food Quality and Preference 30 (2013) 320–327 323 side by side (experiment a). Colour training 3. 1976. 2001). from 37. Martínez et al.8 (the most yellow) and the chroma ranged from 42. Berns. the orangish from 48. For the final difference threshold. Finally. The error between the last and the penultimate session was 6.1. Finally. The percentage of the positive colour differences perceived by both panels (DV) were plotted against the CIELAB colour differences ((DEÃ ab ) instrumentally measured. In this way. An example of the results is shown in Fig.11. Results and discussion 3. & Paffhausen. 1983). Panel training The objective colour of the training solutions. The final results were expressed as ‘‘visual colour difference’’ (DV) which was calculated as the percentage of positive panellists’ responses (1). both commercial orange juices and their dilutions. Strocka. hue angle data ranged from 58. Snyder. Data analysis All statistical analyses were performed using the the program Statistica 8 for Windows (StatSoft. just one panellist was removed from the panel following the criteria used for the selection (Spearman’s ranked correlation coefficient was not significant). eight panellists were selected to be part of the panel. Chroma ranges were 44. The food dye solutions used for tone separation and intensity ranking were analysed by image analysis..01 and finally the reddish serie from 43. respectively). The yellowish serie had a lightness ranging from 56. This mean that some of them was in the group of superior discrimination (scores lower than 16) and others were in the group of average discrimination (scores lower than 100).75 to 0. The software MATLAB R2011b (The MathWorks Inc. each observer in the trained panel (8 panellists) evaluated the 120 pairs of samples once (960 judgments in total). The correlation between the visually perceived and instrumentally measured colour differences were explored following procedures previously described elsewhere (Davidson & Friede.0° to 91. . The severity of the defect can be gauged by the extent of the ‘bulge’.17. and 44.02 to 76.70 to 72.03.65 to 60. 5 shows the samples in the CIELAB space and the coordinate L⁄.29. Kuehni. 3.66 to 0. Brockes. Berns.1.3° in the reddish serie (Fig. Fig. 1998). The visual judgments were made immediately after the instrumental colour measurements in order to avoid the colour variation of the samples. a severe degree of defect would show clear bipolarity with high error scores. respectively.6% in hue and 4. 4). Examples of the results for the Farnsworth-Munsell 100-Hue Test conducted in panellists with (a) average discrimination and (b) superior discrimination.14 to 74. Standard deviations decreased from the first to the last session (from 1. 2001..R.1° to 95. Qiao. the orangish and the reddish series.79. After the panellists sorted the samples out and ranked them according to increasing intensity.57–62.9° in yellowish serie.2.1. 6 shows the evolution in the scores for the sample COJ I along the sessions. similarly did the group of 8 untrained observers (1200 judgments in total). Panellists were trained in different sessions until a consistent panel of assessors was obtained. the 50% of positive responses by the observers was consider as the typical measurement of tolerance or acceptability of colour differences perceived (Alman. The values of the coordinate L⁄ ranged from 64.. Fig. Fernández-Vázquez et al. until maximising the value of r2 (Martínez et al. Then. mild cases with good colour discrimination would show no ‘bulge’ and cannot be identified by this test. 1991. 3.1. 3. Reniff. & Larsen.55 and from 1.8° to 83. 1989. 2.. Fig. Berns et al.

(For interpretation of the references to colour in this figure legend. Representation of the colour coordinates of the commercial orange juice samples used in the training sessions of the panel (a) in the a⁄b⁄ plane and (b) lightness values.002 0. the sensory score were compared to the instrumental values.540 0.308 0.159 0. However. Surprisingly. sample effects were obtained in both tests. / Food Quality and Preference 30 (2013) 320–327 100 90 80 70 h ab 60 50 40 30 20 Ye llowish O rangish Re ddish 8 7 6 5 4 3 2 1 0 Session 1 Session 2 Session 3 Session 4 Session 5 Hue Intensity Fig. were not too pronounced (for example among COJ II and COJ II 50%.242 A three factor ANOVA (assessor. In accordance. Statistically significant correlation coefficient were obtained for both expressions (r⁄ = 0. Hue corresponding to the three series of colourant dilutions (yellowish. in this study colour intensity is proposed as an attribute related to both lightness and chroma. it is important to highlight its uniformity and also to evaluate the visual and instrumental correlations.252 Intensity 0.073 0. not comparing between them. Evolution of the evaluations done by the panel for hue and intensity (mean and standard deviation) along the training sessions. This could be explained by the fact that they evaluated the samples separately.) Table 1 Results of the three factor ANOVA analysis for the colour attributes hue and intensity evaluated by the trained panellist. Furthermore. 5.8 CIELAB units).479 0. lightness and chroma. In order to ascertain if the panellists were correctly trained in colour evaluation. Fig. Equations à à ⁄ ⁄ ⁄ including C à ab and L . differences among these samples in hue. from samples COJ II to COJ II 50% panellists did not find significant differences in hue and intensity. resulting r⁄ statistically significant (r⁄ = À0. respectively). Table 2 shows the final scores given by the panel to the samples evaluated and their standard deviations.74 respectively).245 0. 62 60 58 56 54 b* 52 50 48 46 44 42 40 -4 -2 0 2 4 a* 6 8 10 COJ I COJ II COJ II (6%) COJ II (10%) COJ II (30%) COJ II (50%) COJ II (60%) COJ II (80%) L* 73 72 71 70 69 68 67 66 65 64 12 14 1 (a) (b) Fig. Fernández-Vázquez et al.89 and r⁄ = 0.724 0. 4. and a two factor ANOVA (assessor and sample) for all the samples were performed. as it was the objective (Table 1).05) were obtained for assessors and repetition. Effect Level of Significance Hue Assessor Sample Repetition Assessor – sample Repetition – assessor Repetition – sample 0. sample.281 0. orangish and reddish) used in the selection of the panellists to take part in the training sessions. Non-significant effects (p > 0.324 R.87 and r⁄ = 0. difference in lightness was 1. which supposed an extra difficulty in the evaluator’s task.97). correlation coefficients between intensity and both lightness and chroma were explored. In order to probe the reliability of the panel. 6. resulting significant in both cases (r⁄ = À0. like C ab (100ÀL ) and C ab /L were explored in a try to relate both parameters in a unique correlation with the intensity. however. and repetition) for samples COJ I and COJ II. the reader is referred to the web version of this article.009 0.83. . The correlation between hue and the hab parameter was explored. As it was explained previously. among samples whose differences in colorimetric parameter measured instrumentally were higher (such us COJ II and COJ II 30%) panellists did find significant differences in hue and intensity in their visual evaluations.

used in the training sessions.84 ± 0. CIELAB colour differences DEÃ ab were calculated in each pair of orange juice and ranged from 0.55b 0. . Fernández-Vázquez et al.10.75 ± 0. the observers frequently interpret chroma and lightness as the quantitative expressions of colour.53b 1. Range for the coordinate C Ã ab was 54. Colour differences thresholds In Fig. Different superscripts within columns indicate statistically significant differences (p < 0.6° (the most reddish OJ) to 83.53 ± 0.02 ± 0.32 (the darkest) to 66. Trained panellists ability to perceive colour differences when presented with pairs of orange juices.88b ⁄ way to ‘darkness’ (C Ã ab (100ÀL )). Hue ranged from 77. the CIELAB colour space (a⁄b⁄ plane) (a) and L⁄ (b) illustrates the colour of the samples included in this study and measured by digital image analyses. the reader is referred to the web version of this article.62b Intensity 4. In this sense.43 ± 0.01c 4. The results of the correlation between the instrumental measurements and the visual evaluations done by the trained and untrained panels are shown in Figs.58a 6.39 ± 0.84c 6.65 ± 1.90 ± 0.45c 6.19 ± 1.63a 2.02c 6. Sample pairs were presented to panellist in a side-by-side (a) or separated (b) condition and each point represents the number of panellists (as a percentage) who perceived a difference (DV) between two juices that differ in objective colour measurement as defined by DEÃ ab . The values of the coordinate L⁄ ranged from 61.53. It can be observed that they encompassed a wide range of colour in commercial orange juices. a significant (r⁄ = 0. Samples COJ COJ COJ COJ COJ COJ COJ COJ a–c 325 Hue 4.05). 7. Representation of the colour coordinates of the orange juice samples used for the colour threshold study (a) in the a⁄b⁄ plane and (b) lightness values. / Food Quality and Preference 30 (2013) 320–327 Table 2 Final scores (mean and standard deviation) for hue and intensity given by the panel for the different commercial orange juices (COJI and COJII) and the corresponding dilutions.07c 5. In visual analysis. it is noteworthy to mention that the correlations found between perceived intensity and the proposed equations (which relate ⁄ quantity of colour to the inverse of lightness (C Ã ab /L ) or in a similar 60 59 58 57 68 67 66 65 L* 64 63 62 61 5 6 7 8 9 10 11 12 b* 56 55 54 53 52 60 1 a* (a) (b) Fig. 8 and 9.00 ± 0. 7.) Fig.66 ± 1.93c 4.34a 0.03c 5.10 ± 0.19 ± 1. attach great relevance to establish correlations between the instrumental and sensory evaluation of colour. (For interpretation of the references to colour in this figure legend. 8.26 ± 0.93) correlation between intensity measured by the panel and the orange juice dilutions was found.88 ± 0.2.7° (the most yellowish).87 (the lightest).60c 5.R.23–59.47 to 8.55a 5. I II II II II II II II (80%) (60%) (50%) (30%) (10%) (6%) 3. Finally.91 ± 0. respectively.

. Both experiments (a) side by side and (b) separated samples are presented as separated graphs. and over 1.6 B 2. Special thanks to Francisco José Rodríguez Pulido and Jose Miguel Hernández Hierro for their collaboration on the use of software MATLAB R2011b. 4.. 1986. Table 3 Coefficients for the fitted equations (DV = A/[1 + exp (B + CDEÃ ab )]) resulting from each of the thresholds experiments (side by side and separated samples) in both panels (trained and untrained). 139–151.60 C À0. These results are in accordance with previously published studies in red wine which reported a value around 3. Brown. Junta de Andalucía by the project P08-AGR03784. Both values are very similar (only show a difference of 0.5 À1. H. Moreover.49 0.93 À0. A.326 R. S. Wyszecki & Fielder. with good correlations with the instrumental colour parameters. while over 5 CIELAB units could be refer as ‘big colour differences’ (Berns et al. Poza. Hita. 1971. Luo & Rigg.40 The CIELAB colour-differences instrumentally measured for each pair of samples were plotted against their visually perceived colour differences (DV). 14. Color Research & Application.70 r2 0. Intensity is proposed as a new attribute to jointly evaluate chroma and lightness. 1957. Berns et al. a lower colour difference of 1. Brown & MacAdam.26 r2 0. & Larsen.27 CIELAB units for the separated observations (Fig. Conclusions To sum up.38 1. Melgosa. The coefficients of the fitted curves corresponding to each experiment are shown in Table 3.63 CIELAB units for the side by side experiment and 1. Performance testing of color-difference metrics using a color tolerance dataset.7 B 2. Since this would be the case of the main potential consumer this is an interesting result for the citrus industry. The values of DEÃ ab corresponding to 50% of colour differences perceived (DV) by the trained panel were 1. Alman. However. 1990. Junta de Andalucía. Acknowledgments This work was supported by funding from the Consejería de Innovación Ciencia y Empresa.0 CIELAB units (Martínez et al. & Berns. Furthermore. / Food Quality and Preference 30 (2013) 320–327 Fig.. 1949. 1986.5 CIELAB units for a trained panel and 2. Considering the results previously discussed we propose a value of around 2.59 1. Cheung & Rigg. 8). Previous studies based on standard propose to consider a range of 0. Trained panel A (a) Side by side (b) Separated 93. demonstrating its utility and efficiency. Untrained panellists ability to perceive colour differences when presented with pairs of orange juices. Snyder.5 CIELAB units could be proposed for a trained panel. According to these results a suprathreshold of 1. Up to now. Qiao et al.80 C À1. It might be because the distance between the samples (15 cm) was not enough to change the panellists’ perception comparing separated and side by side evaluation. (1989). D.77 101. D. G. Sample pairs were presented to panellist in a side-by-side (a) or separated (b) condition and each point represents the number of panellists (as a percentage) who perceived a difference (DV) between two juices that differ in objective colour measurement as defined by DEÃ ab . 9). For the final threshold calculations.. 1991.78 CIELAB units for the side by side and 2.0 97. Witt. Brown & MacAdam. the 50% probability was considered as a typical measurement of tolerance or acceptability of colour differences perceived by the observers (Alman et al. Fernández-Vázquez et al.5 CIELAB units is proposed for a trained panel.. Macadam.75 CIELAB units as the threshold and suprathreshold colour difference.73 CIELAB units. 1997.8 CIELAB units as a preliminary estimate of colour-difference threshold in orange juice’s consumers. The higher threshold in the case of side by side samples could be related to an increase in the sensibility of the panel due to the closeness of the samples. Berns.51 Untrained panel A 99. However. As mentioned above.44 CIELAB units in the separated observations experiment (Fig. References Alman.. 1998).. W..38–0. RFV holds a grant from the Consejería de Innovación Ciencia y Empresa. respectively. no experiments on colour-differences perceived by observers in real samples of orange juice have been reported. in the untrained panellists the values of DEÃ ab corresponding to 50% of colour differences perceived were 2. R.19 CIELAB units). the equation of the fitted curve for all the situations was DV = A/[1 + exp (B + CDEÃ ab )]. in this study a specific training method for visual evaluation adapted to orange juice was set up.8 CIELAB units for untrained panellists are proposed. 2001).50 0. These higher values could be explained by the lack of training and knowledge about colour theory. 1991.. 1989. 1942. for the first time a colour-difference threshold of 1. 9. training of the panellists did that these small differences were not enough to change their responses. 1949).

Berns. T. R. A. (1949). 11. & Berns. New York.. Instrumental and visual methods for measuring orange juice color.. K. European Food Research and Technology. H. R. & Monteiro. H.. A.. New color-matching ellipses. A. (1967). Fernández-Vázquez. P. D. R. I. BejinesMejías. González-Miret. CIE (2007).. L. J. 16. 148–155. J.. 143–150). F. Journal of Sensory Studies. Color Research & Application... 99–103. 173–175. D. Effects of a slight color variation on consumer acceptance of orange juice. M. A.R. A. O’Mahony.. C. M. Correlation between visual and instrumental colour measurements of orange juice dilutions: Effect of the background. Huertas.. (2012). E. Salvador. Food Science and Technology International.. Heredia. (1942). K.. E. Buera (Eds. & Balonon-Rosen. Preference mapping of commercial orange juices. 22. Sensory and instrumental evaluation of strawberry yogurt color. Journal of the Optical Society of America. L. Petrus. 436–444... Witt. & Montag. Caivano & M. & Burgueño.. J. G. 116–121. Suprathreshold color-difference ellipsoids for surface colors. W. Frata. (1976). 189–199. Color Research & Application.. Tepper. M. C. 47. Kuehni. 308–311. L.. 808–823.. Luo. W.. Journal of the Optical Society of America. Color in food.. Melgosa. Influence of experimental parameters on the evaluation of color-difference ellipsoids. 137–143. & Ramírez. 16. Cheung. 25–42. R. F. 34. I. S.4. Ou. J. (1998). Heredia. / Food Quality and Preference 30 (2013) 320–327 Barron. ISO.. C. Food Research International. (2011). & Fielder. L. Poelman. (2007). R. 1084–1091. Stinco. 327 Macadam. J. R. Fernández-Vázquez et al. D. Journal of Sensory Studies. M. & Buzz Lig. 229–238. New York: CRC Press.. R. Maraulja. 39. D.. 2007. I. R. J. Influence of colour intensity on the perception of colour and sweetness in various fruit-flavoured yoghurts. The American Journal of Psychology. Food Quality and Preference. S. J. Stinco. (1977). & Vicario. & MacAdam.. D. C. D. I. 26.. R. . (1960). L. L. R. Melendez-Martínez. J. 355–364.. 497–500. (2006). 1135–1152. Brown. Journal of Sensory Studies. 439–444. F. (2006). Journal of the Optical Society of America. (1953). E. A. G. 1–10. C. Proceedings of 5th European Conference on Sensory and Consumer Research.] Ref Type: Patent. Color discrimination of twelve observers.. Davidson. B. J. B. et al. (4-10-2001). R. (1986). Meléndez-Martínez.. Taylor & Francis Group. Domingo. 39. H. Luo.. DigiFoodÒ (Análisis de imagen). M.). R. R. In J. & Hutchings. A.. B. Influence of different backgrounds on the instrumental color specification of orange juices. R.. S. B. A. M. Colour-difference ellipsoids for five CIE colour centres. Álvarez. Visual determination of suprathreshold color-difference tolerances using probit analysis. Luo. Optimisation of food expectations using product colour and appearance. (1983). 247–273. Sensory evaluation of food: Statistical methods and procedures. F. D.. Wyszecki. S. H. Wei. M. & Delahunty. Florida State Horticultural Society. 0124683. Brockes. Journal of the Optical Society of America. Color Research & Application. M. Technologigal and psychophysical aspects (pp. A.. D.. Reniff. G. I. (1997). 49–62. Calvo. Journal of the Optical Society of America. 90.. (2005). Hort. J. (1991). 73. (1993).. & Paffhausen. F. 15.. 11. Influence of color on the discrimination of sweetness. M. 8. sourness and overall liking in orange juice. 66. E. L. M. 61. E. M. Melgosa. Meléndez-Martínez. R.... (1957).. Color Research & Application. (2001).. Color measurement in L⁄a⁄b⁄ units from RGB digital images. 23. 302–313. Heredia. M. 22. 7. & Hita. Parametric effects on surface color-difference evaluation at threshold.. Hita. M. Visual determination of hue suprathreshold color-difference tolerances. 11664–2. D. 43. Vicario. Strocka. 23. R. 169–175. J. & Heredia. Note visual and instrumental color evaluation in red wines. Óptica Pura y Aplicada. S. J.. 2. Color Research & Application. M. A. León. version 8. 16. A. sweetness... Martínez. Brown.. (1990). M. M. Y. J. Fernández-Vázquez. Color Research & Application. & Huggart.. Fruit Processing. & Fiszman. Journal of the Optical Society of America. Visual and instrumental evaluation of orange juice color: A consumers’ preference study. & Vicario. Valim. W. Standard Illuminants for Colorimetry. Pérez. R. T. Colour influence on the perception of flavour. R. Berns. M. 8. Giménez. Marcel Dekker ed. M. Chromaticity-discrimination ellipses for surface colours. Hewson. Some reflections and recent international recommendations on color-difference evaluation. SE-01298. Qiao. E. & Rigg. & Friede. A. 213.. J.. M.. G. Snyder. 471–478. M.. Alman... Cui. Food Quality and Preference. D. Poza. & Rigg. Heredia. N.. Fernández-Vázquez. M. A.0 [Computer software]. M. Pérez. J. Bern. M. (2011). Hita. W.. S. I. (1986). Pangborn. 32. G. 297–316. Switzerland. Commission internationale de l’Eclairage.. Melgosa. & León.. Gambaro. J. C. & Negueruela. R. (2006).. R. 11–22. A. F. J. Color-tolerance data and the tentative CIE 1976 L⁄a⁄b⁄ formula. 581–589. A. STATISTICA. M. M. Reniff. B. StatSoft. [No. L. (2001)... Hernanz Vila. The effect of preparation method and typicality of colour on children’s acceptance for vegetables. (2001). British Patent Entitled apparatus and method for measuring colour (DigiEye System). 145–154. (2012). Alman. 185–195. Color aspects of Florida commercial grapefruit juices. Franco. (1971). I. L. Visual sensitivities to combined chromaticity and luminance differences. I. (2012). Yebra. Proceedings of the Florida State Horticultural Society. (1986). pp.. The size of acceptable color differences. (2001). Color Research & Application. C. R.. Huggart. L. Food Quality and Preference. C.. Visual sensitivities to color differences in daylight. Fisk. & Li.